Abstract and Introduction
Abstract
Plasma biomarkers for Alzheimer's disease-related pathologies have undergone rapid developments during the past few years, and there are now well-validated blood tests for amyloid and tau pathology, as well as neurodegeneration and astrocytic activation. To define Alzheimer's disease with biomarkers rather than clinical assessment, we assessed prediction of research-diagnosed disease status using these biomarkers and tested genetic variants associated with the biomarkers that may reflect more accurately the risk of biochemically defined Alzheimer's disease instead of the risk of dementia.
In a cohort of Alzheimer's disease cases [n = 1439, mean age 68 years (standard deviation = 8.2)] and screened controls [n = 508, mean age 82 years (standard deviation = 6.8)], we measured plasma concentrations of the 40 and 42 amino acid-long amyloid-β (Aβ) fragments (Aβ40 and Aβ42, respectively), tau phosphorylated at amino acid 181 (P-tau181), neurofilament light (NfL) and glial fibrillary acidic protein (GFAP) using state-of-the-art Single molecule array (Simoa) technology. We tested the relationships between the biomarkers and Alzheimer's disease genetic risk, age at onset and disease duration. We also conducted a genome-wide association study for association of disease risk genes with these biomarkers.
The prediction accuracy of Alzheimer's disease clinical diagnosis by the combination of all biomarkers, APOE and polygenic risk score reached area under receiver operating characteristic curve (AUC) = 0.81, with the most significant contributors being ε4, Aβ40 or Aβ42, GFAP and NfL. All biomarkers were significantly associated with age in cases and controls (P < 4.3 × 10−5). Concentrations of the Aβ-related biomarkers in plasma were significantly lower in cases compared with controls, whereas other biomarker levels were significantly higher in cases.
In the case-control genome-wide analyses, APOE-ε4 was associated with all biomarkers (P = 0.011−4.78 × 10−8), except NfL. No novel genome-wide significant single nucleotide polymorphisms were found in the case-control design; however, in a case-only analysis, we found two independent genome-wide significant associations between the Aβ42/Aβ40 ratio and WWOX and COPG2 genes.
Disease prediction modelling by the combination of all biomarkers indicates that the variance attributed to P-tau181 is mostly captured by APOE-ε4, whereas Aβ40, Aβ42, GFAP and NfL biomarkers explain additional variation over and above APOE. We identified novel plausible genome wide-significant genes associated with Aβ42/Aβ40 ratio in a sample which is 50 times smaller than current genome-wide association studies in Alzheimer's disease.
Introduction
Alzheimer's disease (AD) is one of the greatest health challenges, affecting tens of millions of people worldwide. The clinical diagnosis of this disease is, however, often inaccurate; around 25% of people with clinical AD do not have underlying pathology at autopsy, and many people who have not yet developed AD-type dementia have incipient pathology, the prevalence of which increases with age.[1] Detecting AD at the earliest possible stage remains essential to combating its effects and to further our understanding of this devastating illness. By diagnosing early, we can better understand how the disease progresses, plan and implement treatments earlier and monitor response to drugs currently being trialled.
Amyloid-β (Aβ) and tau pathology are the defining pathological features of AD.[2] For many years, it has been possible to detect AD pathology (amyloid aggregation, tau tangles and neurodegeneration) using imaging and CSF biomarkers. Although CSF and PET biomarkers of Aβ and tau are highly accurate for detecting disease pathology,[3] the costs, invasive nature and low availability of the tools needed to detect these biomarkers hamper their feasibility for use in clinical diagnostic practice and for screening in clinical trials.
Assays for plasma Aβ fragments [ratio of Aβ1–42 (Aβ42) to Aβ1–40 (Aβ40)] reflect brain amyloidosis;[4–7] however, these assays have limitations, including the impact of substantial peripheral Aβ production.[8] By contrast, CSF and plasma tau phosphorylated at threonine 181 (P-tau181) is a highly specific pathological marker of AD that remains normal in other dementias.[9,10] Glial fibrillary acidic protein (GFAP) and neurofilament light chain (NfL) are putative non-amyloid plasma-based biomarkers indicative of ongoing neuroinflammatory and neurodegenerative disease processes. Increased GFAP suggests abnormal activation and proliferation of astrocytes, for instance secondary to neuronal damage. It has been shown that GFAP levels in plasma and CSF are higher in AD and correlate with cognitive impairment.[11–13] Plasma NfL is a marker of neuronal injury, increased in AD,[14] but this biomarker has low specificity, because increases are also reported in several other neurodegenerative disorders.[13,15,16] Thus, while NfL has potential as a monitoring biomarker, GFAP might be a valuable prognostic biomarker, predicting incident dementia.[13] Recent reports show that plasma P-tau181 concentration starts to increase around 15 years prior to clinical disease onset in familial AD[17] and that plasma P-tau181 predicts disease neuropathology at least 8 years prior to autopsy in sporadic disease.[10]
Early disease prediction can be helped with genetic data as an individual's genetic makeup does not change over time and genetic data are precise and inexpensive to measure; however, the prediction accuracy using genetics is limited.[18] Biomarkers, in contrast to genetics, can only indicate the presence of AD pathology after the disease has already been triggered, i.e. a biomarker change marks the onset of a pathological process. Nevertheless, the prediction accuracy of e.g. P-tau181 and P-tau217 for discriminating AD from other neurodegenerative diseases,[19–21] when combined with APOE genotype, memory and executive function phenotypes, was reported to reach area under receiver operating characteristic curve (AUC) > 90% in predicting the progression from mild cognitive impairment (MCI) to AD in two relatively small samples of participants (n = 340 and 543).[22]
Identifying genetic loci associated with biomarkers could aid understanding of the specific pathophysiological components underpinning these biomarkers. Genome-wide association studies (GWAS) of CSF biomarkers in AD case/control samples have found loci in genes GEMC1 and OSTN[23] as well as more commonly reported loci such as the TREM cluster, APOE, APOC and TOMM40.[24] However, these have also only focused on small sets of biomarkers, typically P-tau181 and Aβ42. GWAS of blood plasma P-tau181 and NfL levels[25,26] have identified only loci within the APOE genomic region, and only for P-tau181. Investigation of the relationship between AD polygenic risk score (PRS) and plasma P-tau181[27] has revealed highly significant associations with PRS containing the APOE region (P = 3 × 10−18−7 × 10−15) and moderate association when APOE was excluded. GWAS studies for plasma Aβ40, Aβ42 and Aβ42/40 ratio in non-demented participants from population-based studies have identified GWAS significant variants in APOE and BACE1 genes, and APP, PSEN2, CCK and ZNF397 genes in gene-based analysis.[28]
The aims of this study were (i) to test the prediction ability of the biomarkers for clinical AD diagnosis in our cohort (over and above commonly used predictors such as APOE, age and AD PRS); and (ii) to identify genetic loci associated with these plasma biomarkers. The latter may shed light on which single nucleotide polymorphisms (SNPs) associated with clinical AD are also associated with plasma biomarkers. This could help to further refine the relevance of the AD GWAS genes to different biological processes, which the biomarkers represent. To that end, we measured plasma biomarkers in a sample of 1439 early and late onset AD cases {mean age 68 years [standard deviation (SD) = 8.0]} and 508 elderly screened controls [mean age 82 years (SD = 6.7)]. We used ultrasensitive Single molecule array (Simoa) assays to measure P-tau181, NfL, GFAP, Aβ40 and Aβ42, and calculated the ratio of Aβ42/40. We then tested these biomarkers for association with the clinical diagnosis of AD and, in case samples, the relationship of the biomarkers with age at sample collection, age at onset and disease duration. To identify genetic loci associated with these biomarkers, we undertook a GWAS for P-tau181, NfL, Aβ40, Aβ42, ratio of Aβ42/40 and GFAP biomarkers in the largest case-control sample set to date.
Brain. 2023;146(2):690-699. © 2023 Oxford University Press